Affordable AI-Powered Exergame for Stroke Rehabilitation and Upper-Limb Function Evaluation

Overview

This study introduces a low-cost, AI-driven exergame that enables simultaneous stroke rehabilitation and automatic evaluation of upper-limb motor function using only a standard camera. The system demonstrated strong correlations with the clinical Fugl-Meyer Assessment and accurately classified motor severity, offering a scalable and interpretable tool for remote stroke monitoring.

Background

Stroke is a leading cause of long-term disability, frequently impairing upper-limb motor function and necessitating ongoing assessment. The Fugl-Meyer Assessment (FMA) is the clinical gold standard for evaluating motor recovery but is time-intensive and requires specialized clinicians. Virtual reality and gaming technologies have shown promise in stroke rehabilitation, yet many solutions rely on costly sensors or lack automated evaluation capabilities. This study addresses these gaps by developing a sensor-free, AI-powered exergame that provides both therapy and objective motor function assessment during gameplay.

Data Highlights

Key Findings

• Sixteen kinematic and spatiotemporal features extracted from 2D hand and arm trajectories strongly correlated with FMA scores.
• Features such as hand angle, range of motion, movement area, traveled distance, and shoulder–elbow coordination effectively stratified motor severity.
• A lightweight linear regression model predicted FMA scores with high accuracy (Spearman ρ = 0.92, R² = 0.89, RMSE = 4.42).
• Severity classification achieved 86–93% accuracy, outperforming more complex machine learning models.
• The approach is sensor-free, using only a standard camera and the MediaPipe framework, enhancing scalability and accessibility.
• The system provides immediate feedback during gameplay, reducing clinical workload and enabling telerehabilitation and remote monitoring.

Clinical Implications

This AI-powered exergame offers a practical, low-cost solution for continuous upper-limb rehabilitation and motor function assessment post-stroke without the need for specialized equipment or personnel. Its high predictive accuracy and interpretability support its integration into clinical workflows and remote monitoring programs, potentially improving patient access to rehabilitation and enabling timely adjustments to therapy.

Conclusion

The proposed AI-driven exergame represents a scalable, interpretable, and effective tool for simultaneous stroke rehabilitation and objective upper-limb function evaluation. Its sensor-free design and strong correlation with clinical standards position it as a promising approach for enhancing post-stroke care and telerehabilitation.

References

1. Tsao, C. W. et al. 2022 -- Heart Disease and Stroke Statistics-2022 Update: a report from the American Heart Association
2. Feigin, V. et al. 2018 -- Global, regional, and country-specific lifetime risks of stroke, 1990 and 2016
3. Saposnik, G. & Levin, M. 2011 -- Virtual reality in stroke rehabilitation: a meta-analysis and implications for clinicians
4. Pyae, A., Luimula, M. & Smed, J. 2015 -- Rehabilitative games for stroke patients
5. Thomson, K., Pollock, A., Bugge, C. & Brady, M. C. 2016 -- Commercial gaming devices for stroke upper limb rehabilitation: a survey of current practice